Add crud storage metrics

[ArtDu]

We have a problem finding queries that parse a lot of tuples on storage.

• It could be a problem if query is too long (netbox timeout raise error and router will get nothing if batch_size huge or not certain)
• Or there a lot of tuples are looked up on storage, but return a few on router

Reproducer with netbox timeout:

1. Create space with non unique index (bad_index)
2. Insert ~1_000_000 tuples with the same field value (bad_value) for non unique index (bad_index)
3. Start fiber with requests by pk
4. Run request with condition where bad_index key == bad_value